Performance portable implementation of a Poisson equation solver

IIT Goa have developed unsteady Poisson Equation solver code, which is a model template for our in-house large-scale parallel turbulent flow solver, it uses fully explicit time integration and local spatial discretization; hence, it is highly parallelizable on modern GPU architectures. As this CUDA implementation is limited to NVIDIA GPUs and we also look for other alternatives, we migrated this CUDA code to SYCL using Intel® DPC++ Compatibility Tool. Migrated SYCL code is capable to run on heterogenous architecture of NVIDIA as well as Intel® GPU. Using Intel® VTune Profiler, we optimized SYCL code on Intel® Data Center GPU Max Series.